Hip-knee collaborative mapping method for self-adaptive speed change of dynamic thigh prosthesis

Abstract

The invention belongs to the field of power artificial limb control, and particularly relates to a hip-knee collaborative mapping method for self-adaptive speed change of a power thigh artificial limb. A normalized translational motion lagged hip joint angle is used as an input, a normalized knee joint angle is used as an output, a mapping relation between the hip and the knee joint is established, the mapping provided by the invention can realize continuous prediction of the knee joint angle in a variable speed scene, and phase splitting is not needed; the method has the advantages of low calculation complexity and small sensing requirement, and is very beneficial to real-time control of the artificial limb; and meanwhile, the mapping parameters do not need to be adjusted for different wearers, so that the workload of parameter adjustment is greatly reduced. According to the method, the synergy of the gait of the artificial limb and the gait of the healthy limb is improved, and a brand-new, simple and effective method is provided for artificial limb control.

Description

Technical Field

[0001] The invention belongs to the technical field of power prosthesis control, and in particular relates to a hip-knee coordinated mapping method for adaptive speed change of a power thigh prosthesis. Background Art

[0002] Powered thigh prostheses are important devices for thigh amputees to restore their mobility. In current research, the most commonly used method for controlling powered prostheses is impedance control based on finite state machines (FSMs). This method divides the entire gait cycle into 4 to 5 stages based on joint angle, angular velocity, or ground reaction force criteria, and adjusts the impedance parameters for each stage. Although this method is relatively simple to deploy, it usually involves 20 to 30 control parameters and requires several hours of parameter tuning. In addition, different finite state machines must be designed for different gait modes (such as variable speed or variable slope). Even if the workload of parameter adjustment for a specific gait stage or mode can be reduced through reinforcement learning or bio-inspired parameter selection methods, control parameters still need to switch between different gait stages and modes. Once the gait stage is misidentified, incorrect movement of the prosthesis sometimes causes the amputee to fall. Therefore, trying to find a continuous and adaptive control method is a research goal of those skilled in the art.

[0003] One method to achieve continuous control is to construct a phase variable that increases monotonically with the time of the entire gait cycle. This phase variable can replace time. By expressing the preset joint trajectory as a function of the phase variable, continuous position control of the entire cycle can be achieved. The most common phase variable is the polar coordinate angle of the phase diagram of the hip joint angle and angular velocity. This method can enable amputees to walk at different speeds using powered knee-ankle prostheses by adjusting only six control parameters. However, the continuous phase variable must be guaranteed to increase monotonically, and the moment when the thigh angular velocity is small makes this condition invalid, which leads to the need to segment the phase variable process, making the control process not completely continuous. If a direct mapping relationship between the hip joint and the knee joint can be directly established, the construction of the phase variable can be omitted, making the control of the prosthetic knee joint simpler and more continuous. Summary of the invention

[0004] The purpose of the present invention is to propose a hip-knee collaborative mapping method for adaptive speed change of a powered thigh prosthesis, aiming to solve the problems of discontinuity and inability to adaptively change speed of current prosthetic motion control.

[0005] The present invention proposes a hip-knee collaborative mapping method for adaptive speed change of a powered thigh prosthesis, which uses the normalized translational motion time-delayed hip joint angle as input and the normalized knee joint angle as output to establish a mapping relationship between the hip and knee joints.

[0006] The invention provides a hip-knee collaborative mapping method for adaptive speed change of a powered thigh prosthesis. The mapping method is implemented by an embedded system and a prosthesis collaborative control framework. The embedded system includes an IMU, an AD7606 isolation module, a signal amplifier, a six-dimensional force sensor, a joint encoder and an STM32 microprocessor. The IMU is connected to the STM32 microprocessor through a CAN port, the output end of the six-dimensional force sensor is connected to the input end of the signal amplifier, the output end of the signal amplifier port is connected to the input end of the AD7606 isolation module, and the output end of the AD7606 isolation module is connected to the STM32 microprocessor through an SPI interface; the output end of the joint encoder is connected to an Elmo driver, the Elmo driver is driven by a motor, and a lithium battery is respectively connected to the motor and the STM32 microprocessor through a voltage divider; the STM32 microprocessor is connected to a computer; the prosthesis collaborative control framework consists of a collaborative mapping generator and a PD torque controller; the specific steps are as follows:

[0007] (1) obtaining a motion trajectory of a healthy user walking at different walking speeds; the motion trajectory includes: a hip joint angle and a knee joint angle;

[0008] (2) Establishing a basic model for hip-to-knee mapping; the basic model is:

[0009]

[0010] C m (v) = B 0m +B 1m v,

[0011] in: is the knee joint angle normalized to between -1 and 1 according to the maximum and minimum values ​​of the knee joint, θ hip (v, t-τ) is the hip joint angle normalized to between -1 and 1 according to the maximum and minimum values ​​of the hip joint, τ is the motion lag parameter, v is the walking speed, the unit is km / h; the maximum and minimum values ​​of the knee joint are set to 65° and -5°; the maximum and minimum values ​​of the hip joint are set to 35° and -20°; the motion lag parameter is the time difference that causes the hip joint angle to shift to the right on the time axis;

[0012] (3) Searching for the motion lag value by traversing; in a discrete time series, the value of the number of time points traversed from 0 to one gait cycle is a candidate value of the motion lag, so that the hip joint angle sequence is translated on the time axis by the time difference of the candidate motion lag, and the specific value of the coefficient to be determined in the basic model is obtained by the least square method, and the mapped knee joint angle is obtained, and the mapped angle and the real knee joint angle are used to make a Pearson correlation coefficient. The above process is repeated to obtain a one-to-one correspondence between the correlation coefficient and the candidate motion lag, and the candidate motion lag that maximizes the correlation coefficient is taken as the final motion lag value;

[0013] (4) Fitting a linear model of motion delay, gait speed and gait cycle; the linear model has the following specific form:

[0014] τ=A0+A1vT+A2T, where v is the walking speed and T is the duration of the gait cycle;

[0015] The motion time lag corresponding to the gait under each of the above-mentioned gait speeds and gait cycles is fitted by least squares to obtain specific values ​​of A0, A1, and A2; the linear model plays a role in online estimation of motion time lag;

[0016] (5) The basic model is deployed on the embedded system; the IMU corresponds to the hip joint angle, the AD7606 isolation module, the signal amplifier and the six-dimensional force sensor correspond to the ankle joint, the joint encoder corresponds to the knee and ankle joints, and the control frequency of the STM32 microprocessor is set to 100 Hz. In each control cycle, the STM32 microprocessor first obtains the hip joint angle through the IMU attached to the front of the thigh; then, based on the six-dimensional force sensor data of the ankle joint, it is determined whether the prosthesis has a heel-to-ground action. If so, the linear model described in step (4) is used to estimate the current motion lag based on the duration data of the previous gait cycle and the estimated walking speed data. If there is no lag, the motion lag of the previous control cycle is used as the current motion lag; then the hip joint angle value of the first n cycles of the current control cycle is read as the input of the mapping, and the mapped knee joint angle at the current moment is obtained through the basic model; finally, the mapped knee joint angle is used as the reference angle of the PD control, the control torque is calculated to obtain the PD controller, and the control torque is converted into a current signal and sent to the motor to control the prosthetic knee joint movement.

[0017] The beneficial effects of the present invention are:

[0018] (1) Hip-knee collaborative mapping can achieve continuous prediction of knee joint angle without phase separation, thus achieving continuous control of the prosthetic knee joint;

[0019] (2) Hip-knee collaborative mapping adaptively adjusts the size of motion delay according to the current gait speed and gait cycle duration to achieve adaptive speed change of the prosthetic knee joint;

[0020] (3) The form of hip-knee collaborative mapping is simple, with low sensor requirements and low computational burden on the control board, which can meet the real-time control requirements of the prosthesis;

[0021] (4) Hip-knee collaborative mapping explores the basic laws of human variable speed movement. There is no need to adjust the mapping parameters for different wearers, and there is no need for phase control. Compared with the phase impedance control method, the control parameters are greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Construction of hip-knee synergy maps.

[0023] Figure 2 SolidWorks design and prototype for a powered thigh prosthesis.

[0024] Figure 3 Embedded systems for powered thigh prostheses.

[0025] Figure 4 A collaborative control framework for prostheses based on hip-knee collaborative mapping. DETAILED DESCRIPTION

[0026] The present invention is further described below by way of embodiments in conjunction with the accompanying drawings.

[0027] Embodiment 1:

[0028] like Figure 1 As shown in the figure, the left figure is the original phase diagram of the hip joint and the knee joint, and the right figure is the phase diagram of the hip joint and the knee joint after the translational motion lag. It can be found that after the translational motion lag, the angles of the hip joint and the knee joint have an approximate one-to-one correspondence, so the mapping between the hip joint and the knee joint can be obtained through the basic model. The figure also gives the steps for finding the motion lag. First, the hip joint angle moves backward through the candidate motion lag (the number of time points from 0 to one gait cycle), and then the least squares method is used to complete the mapping from the hip joint angle to the knee joint angle. By maximizing the correlation coefficient between the experimental knee joint angle and the mapped knee joint angle, the value of the motion lag can be determined.

[0029] like Figure 2As shown in Figure 1, its knee and ankle joints are driven by high-torque motors (RI80 KV75, T-motor, China), with a range of motion of -5° to 120° at the knee joint and a continuous torque of ~110Nm, and a range of motion of -35° to 35° at the ankle joint and a continuous torque of ~90N-m. The rated torque can meet the needs of daily walking activities. The knee joint is driven by a harmonic drive (CAD-20-50-2A-GR, harmonic drive, Japan) with a transmission ratio of 50, while the ankle joint is driven by a ball screw connecting rod with a diameter of 10 mm and 5 mm. The prosthesis weighs about 4.5 kg, and most of the components are made of 7075-T6 aluminum alloy, with a few shafts and bearings made of 45 steel.

[0030] like Figure 3 As shown in the figure, the system includes a power system, a control unit and a sensor system. A 48V / 20ah lithium battery powers the microprocessor and the motor through a cable, and a voltage divider is used to distribute 24V voltage to the microprocessor and 48V voltage to the motor. The control unit integrates an STM32F407 microcontroller (STM32F4, ST, Switzerland), two controllers (G-MOLTWIA50 / 100EEOT, Elmo, Israel) and two high-torque motors. The microprocessor completes all calculations and sends the current command to the controller through the CAN port, and the controller controls the motor in a three-phase current manner. During real-time control, the microprocessor can transmit data wirelessly to a laptop through the serial port. The sensor system consists of an IMU (MTi-630, Xsens, the Netherlands), a six-axis force sensor (M3713D, Sunrise Instruments, China) and two incremental joint encoders (RLS, RM44D01, Slovenia). Among them, IMU is used for hip angle measurement, six-dimensional force sensor is used for ankle joint force measurement, and joint encoder is used for knee and ankle motor angle measurement. These three transmit data to the control unit for processing through CAN, SPI interface and feedback cable respectively.

[0031] like Figure 4 As shown, the control framework consists of a collaborative mapping generator and PD torque control. In the collaborative mapping generator, the prosthesis estimates the gait cycle and walking speed according to the real-time IMU and six-dimensional force data, and then obtains the motion time delay according to the linear model. The reference knee joint angle is obtained according to the collaborative mapping, and the above reference angle is input to the PD controller to control the prosthesis movement.

[0032] Take a subject with a thigh amputation as an example. The subject is 173cm tall, weighs 75kg, is 49 years old, has worn a prosthesis for 20 years, and the residual limb accounts for 40% of the thigh length. The subject performed three sets of experiments, each of which was completed using a collaborative control method and a phase impedance control method, the latter being a control experiment for the former. The first set of experiments was five round trips on a 3-meter sidewalk without handrails; the second set of experiments was walking on a treadmill at 2.0, 2.5, 3.0, 3.5, 4.0 and 4.5 km / h for one minute each; the third set of experiments was walking on a treadmill with continuous variable speed, the treadmill speed increased from 2.0 km / h to 4.5 km / h at 0.1 km / h per second, and then decreased from 4.5 km / h to 2.0 km / h at the same speed. During the experiment, the motion capture system (Vicon, Oxford, UK) recorded the lower limb movement data throughout the process. STM32 completes the real-time reception, processing and transmission of all data in the control process, and sends the data to the computer through the serial port at a frequency of 100Hz for data analysis after the experiment. It should be noted that before each experiment, the subject lifted the prosthesis upright off the ground for 5 seconds to calibrate the initial angle of the IMU and the initial value of the ankle joint six-dimensional force to 0.

[0033] After a brief preliminary adjustment, we set the maximum and minimum values ​​of the subjects' hip joints to 35° and -25°, the maximum and minimum values ​​of the subjects' knee joints to 70° and -5°, and the K p and K d The control parameters were set to 2.2 and 0.25, and A0 was adjusted from 0.25 to 0.24. Apart from these fine-tuning, the other parameters did not need to be adjusted and remained unchanged throughout the experiment. The experimental results show that in Experiment 1, the subjects could complete five round trips on the sidewalk at a comfortable speed; in Experiment 2, the subjects used collaborative control to better capture the changes in lower limb movement. As the speed increased, the prosthetic knee joint showed a larger knee swing amplitude and frequency. In contrast, the phase-splitting impedance control could not adapt to the changes in knee swing amplitude and frequency with the walking speed, and there was also a misidentification of the gait phase; in Experiment 3, the collaborative control also achieved the synchronous increase of the amplitude and frequency of the knee and hip joint movements with the walking speed, showing good coordination, while the phase-splitting impedance control made the subjects feel obviously strenuous when changing speed. Therefore, under this control, the knee joint could not adjust the amplitude, resulting in the hip joint needing to increase the amplitude additionally to adapt to the changes in walking speed, consuming more physical strength. In general, the prosthesis collaborative control method based on hip-knee collaborative mapping can reduce the compensatory behavior of the hip joint, achieve the coordination between the knee and hip joints, greatly reduce the control parameters and adjustment time, and realize continuous adaptive speed change of the prosthesis.

Claims

1. A hip-knee collaborative mapping method for adaptive speed change of powered thigh prosthesis, characterized in that The normalized translational motion time-delayed hip joint angle is used as input, and the normalized knee joint angle is used as output to establish a mapping relationship between the hip and knee joints; the mapping method is implemented by an embedded system and a prosthesis collaborative control framework, wherein the embedded system includes an IMU, an AD7606 isolation module, a signal amplifier, a six-dimensional force sensor, a joint encoder and an STM32 microprocessor, the IMU is connected to the STM32 microprocessor through a CAN port, the output end of the six-dimensional force sensor is connected to the input end of the signal amplifier, the output end of the signal amplifier port is connected to the input end of the AD7606 isolation module, and the output end of the AD7606 isolation module is connected to the STM32 microprocessor through an SPI interface; the output end of the joint encoder is connected to an Elmo driver, the Elmo driver is driven by a motor, and a lithium battery is connected to the motor and the STM32 microprocessor respectively through a voltage divider; the STM32 microprocessor is connected to a computer; the prosthesis collaborative control framework consists of a collaborative mapping generator and a PD torque controller; the specific steps are as follows: (1) obtaining a motion trajectory of a healthy user walking at different walking speeds; the motion trajectory includes: a hip joint angle and a knee joint angle; (2) Establishing a basic model for hip-to-knee mapping; the basic model is: C m (v)=B 0m +B 1m v, in: is the knee joint angle normalized to between -1 and 1 according to the maximum and minimum values ​​of the knee joint, θ hip (v, t-τ) is the hip joint angle normalized to between -1 and 1 according to the maximum and minimum values ​​of the hip joint, τ is the motion lag parameter, v is the walking speed, the unit is km / h; the maximum and minimum values ​​of the knee joint are set to 65° and -5°; the maximum and minimum values ​​of the hip joint are set to 35° and -20°; the motion lag parameter is the time difference that causes the hip joint angle to shift to the right on the time axis; (3) Searching for the motion lag value by traversing; in a discrete time series, the value of the number of time points traversed from 0 to one gait cycle is a candidate value of the motion lag, so that the hip joint angle sequence is translated on the time axis by the time difference of the candidate motion lag, and the specific value of the coefficient to be determined in the basic model is obtained by the least square method, and the mapped knee joint angle is obtained, and the mapped angle and the real knee joint angle are used to make a Pearson correlation coefficient. The above process is repeated to obtain a one-to-one correspondence between the correlation coefficient and the candidate motion lag, and the candidate motion lag that maximizes the correlation coefficient is taken as the final motion lag value; (4) Fitting a linear model of motion delay, gait speed and gait cycle; the linear model has the following specific form: τ=A0+A1vT+A2T, where v is the walking speed and T is the duration of the gait cycle; The motion time lag corresponding to the gait under each of the above-mentioned gait speeds and gait cycles is fitted by least squares to obtain specific values ​​of A0, A1, and A2; the linear model plays a role in online estimation of motion time lag; (5) The basic model is deployed on the embedded system; the IMU corresponds to the hip joint angle, the AD7606 isolation module, the signal amplifier and the six-dimensional force sensor correspond to the ankle joint, the joint encoder corresponds to the knee and ankle joints, and the control frequency of the STM32 microprocessor is set to 100 Hz. In each control cycle, the STM32 microprocessor first obtains the hip joint angle through the IMU attached to the front of the thigh; then, based on the six-dimensional force sensor data of the ankle joint, it is determined whether the prosthesis has a heel-to-ground action. If so, the linear model described in step (4) is used to estimate the current motion lag based on the duration data of the previous gait cycle and the estimated walking speed data. If there is no lag, the motion lag of the previous control cycle is used as the current motion lag; then the hip joint angle value of the first n cycles of the current control cycle is read as the input of the mapping, and the mapped knee joint angle at the current moment is obtained through the basic model; finally, the mapped knee joint angle is used as the reference angle of the PD control, the control torque is calculated to obtain the PD controller, and the control torque is converted into a current signal and sent to the motor to control the prosthetic knee joint movement.

Images